WO1993008666A1

WO1993008666A1 - Method for the non-hierarchical routing of traffic in a communications net

Info

Publication number: WO1993008666A1
Application number: PCT/EP1992/002302
Authority: WO
Inventors: Harro Lothar Hartmann
Original assignee: Siemens Aktiengesellschaft
Priority date: 1991-10-15
Filing date: 1992-10-06
Publication date: 1993-04-29
Also published as: DE59207963D1; CA2121240C; ATE148292T1; JPH06510645A; EP0608279B1; EP0608279A1; US5537468A; JP2851432B2; CA2121240A1; ES2098544T3

Abstract

Prior art non-hierarchical methods for the routing of communications traffic give optimum throughputs only under certain traffic loads. In order to ensure optimum throughtputs under all load conditions, the invention proposes that the number of alternative routes available for overflow traffic be modified in near real time as a function of the traffic load conditions on the alternative routes.

Description

Method for non-hierarchical traffic routing in a communication network

The invention relates to a method for non-hierarchical traffic control in a communication network according to the preamble of claim 1.

A method according to the preamble of claim 1 is already known from ITC-11 (1985) pp. 795-801, "use of a trunk status map for real-ti e DNHR" by G.R. Ash known. The known method has significant throughput losses when overloaded.

All previously known dynamic methods for non-hierarchical traffic control achieve optimal throughput values only in certain traffic load ranges, i.e. they show significant losses in throughput, either in the event of an overload or in the event of a high or planning load.

The invention is based on the problem of achieving optimum throughput values over all traffic load ranges.

This problem is solved by the features of claim 1.

Through the real-time, state-controlled adaptation of the number and sequence of alternative routes available for traffic, i.e. the alternative route sequence, almost optimal throughput values or minimal transport costs are achieved in all load ranges.

If all alternative routes of a traffic pair are overloaded, only the planned routes or the planned route are finally available for traffic. This state-controlled route selection on the one hand reduces the traffic pair-specific high load, but at the same time favors the background load (traffic via planned routes plus overflow from other network sections) on the affected link sections. Therefore, the successful throughput of the entire communication network increases. The method according to the invention for non-hierarchical traffic routing therefore works in a conservative reservation and is compatible with other routing environments. It uses the generally non-coinciding traffic minima (Multi Hour Routing) in real time by knowingly assigning free channels (Multi Service Routing). -

The method according to the invention can be referred to as a state-controlled dynamic non-hierarchical routing method (state-controlled dynamic non-hierarchical routing, SDNHR for short).

An embodiment of the method according to the invention has the particular advantage that the traffic load on the signaling channels is reduced compared to a central routing method and the throughput is thus promoted. This is particularly important in the case of a communication network with many switching nodes and a large number of alternative routes. Otherwise, the processing load is distributed to the local processors. As a result, security-relevant, real-time efficient routing methods with decentralized "Processing Load Sharing" can be implemented.

A further embodiment of the method according to the invention has the particular advantage that the originating switching node can easily classify the traffic load status of an alternative route. A further embodiment of the method according to the invention has the particular advantage that the originating switching node can classify the traffic load state of an alternative route, taking into account the subsequent link loads.

A further embodiment of the method according to the invention has the particular advantage that the originating switching node can classify the traffic load status of an alternative route, taking into account the status of the follow-up links, without having to use a method for estimating the blocking probability of an alternative route.

A further embodiment of the method according to the invention has the particular advantage that the alternative way traffic (overflow traffic) is displaced even more in favor of the planned traffic in the event of an overload.

The figure list of the drawing follows.

1 shows a communication network in which the method according to the invention is used.

2 shows the processing structure present in a switching node for realizing the method according to the invention.

3 illustrates the throughput optimization in the case of adaptive path fan control.

4 shows the reduction in the processing load ratio (PRLR) in various embodiments of the method according to the invention.

A special embodiment of the method according to the invention is explained in more detail below with reference to the drawing. 1 shows a communication network which includes switching nodes A, B, C, D, E and F. Each of the switching nodes A to F shown works with a route plan specified by the network operator or a central network management, in which, in addition to a planned route (preferably direct route) or a plurality of planned routes, a centrally specified number M of alternative routes to each connection destination or destination switching nodes are included. In the case of several planned routes, the sequence of these planned routes is predetermined by the central network engineering or network management. In exceptional network situations, the number and sequence of alternative routes can also be specified by the central network management, for example in 5-minute intervals. In normal network situations and therefore in dominant operating time intervals, the alternative routes are potential real-time routes. According to the invention, these alternative routes are assigned by the routing process in a decentralized and state-controlled manner to the connection requests (calls) arriving in a switching node in accordance with an alternative route sequence determined by the routing process in relatively short cycle times. 1 further shows a first switching center VI, which is connected to the non-hierarchical communication network via the switching node A, and a second switching center V2, which is connected to the non-hierarchical communication network via the switching node D. Since the connection of the two switching centers VI and V2 is fixed or hierarchical in terms of planning, the connection shown in FIG. 1 takes place between the switching centers VI and V2 via the switching nodes A and D of the non-hierarchical communication network. The incoming calls from the switching center VI to the originating switching node A can be switched via alternative connection paths to the destination switching node D, which is shown in FIG. 1 by uninterrupted connection arrows. The in FIG Connection paths shown in dashed lines between the originating switching node A and the destination switching node D are connection paths that were not selected by the routing process, ie were not included in the alternative route sequence in the update update interval under consideration.

2 shows the processing structure present in the selected originating switching node A for implementing the method according to the invention. The implementation of the invention, which is explained in more detail below, generally includes software and hardware components with coordinated task assignments.

The processing structure shown in FIG. 2 comprises a routing process RP, a routing table RT (D) and a local trunk state table LTSM. The routing process RP is used to determine an alternative route sequence AWS within the centrally permitted set of M alternative routes. In doing so, he uses the local trunk status table LTSM, in which the status of the links of all alternative routes are stored, and possibly contains a binary statement about the availability of transit nodes TN of the alternative routes.

The routing table RT (D) contains a route sequence SEQ of connection routes VW, by means of which the switching process in the present example is provided with a planning direct route D in the first place and a current alternative route AW in the second place. The current alternative route AW is made available to the switching process until it is occupied by the overflow traffic or a blockage occurs. There is then a message to the routing process RP, whereupon the next alternative route according to the alternative route sequence AWS as the new current alternative route instead of the previous one writes current alternative route into the routing table.

The local trunk state table LTSM contains the states of all centrally approved alternative routes of a respective origin-destination switching node pair, here e.g. of the switching node pair A-D. The routing process uses this to determine the alternative route sequence within the update interval, i.e. the order and scope of the alternative routes used for the overflow traffic. All incoming calls in the next update interval use this alternative route sequence in sequential overflow.

The special origin-destination switching node pair A-D is considered below.

The first connecting route is the route D, which is specifically a direct route. The routing process specifies the least stressed connection path via the switching node E as the first alternative path AW. The switching node E here represents a transit node TN which is available for the traffic. The availability of a transistor node TN is indicated in the local trunk status table by an in

Parentheses behind the name of the transit node are the binary value. The first alternative route runs via a first link L1 (A), which according to the local trunk status table has 12 free channels, and via a second link L2 (D), which has 18 free channels. Incoming multichannel connection requests of up to 12 channels can therefore also be offered in this alternative way.

The time interval after which the local trunk status table updates with non-local status information (update interval dT) is about 10 see. This ensures on the one hand a real-time recording of the occupancy status of the alternative routes, and on the other hand the frequency of polling of the local trunk status table and thus the evaluation load is low.

The routing process ensures that the incoming calls to those alternative routes with the largest number of channels available throughout, i.e. with the lowest traffic load. The following process components are required for this:

a) LTSM entries in the grid of the update interval: available channels, minus trunk reservation, minus BIAS estimates. b) Allocation of the incoming calls to the free (most idle) alternative route with overflow to subsequent alternative routes. The ranking is based on continuously decreasing free channel numbers. c) Call blocking if no alternative route is available.

If the second link of the first alternative route is blocked when the connection is being set up, contrary to the statement of the local trunk status table, crankback occurs and the connection route via the transit switching node F is used as the second alternative route. Further alternative routes are not available in the update interval of the local trunk status table under consideration and can therefore not be used for establishing the connection. The connection path via the intermediate switching node B is not selected by the routing process due to the lack of free channels on the first link L1 (A) and the connection path via the intermediate switching node C as a result of an overload in C and is therefore not made available to the switching process. The routing process thus generates a partial order of free connection path capacities from a centrally specified set of M alternative routes. set of alternative routes, where T ≤ M.

In the case of incoming multi-channel calls, the available free connection capacities in the same order of priority are decisive. The assignment of two HO channels, each with 384 kbit / s of required capacity, leads to the alternative route being booked out via the intermediate switching node E during the update interval under consideration, and thus to a lowering of the range or disembarkation of this alternative route in the next update interval.

The number of free trunks of a link specified in the local trunk status table according to FIG. 2 has already been reduced by a certain number of trunks. This artificial reduction takes into account, on the one hand, the usual link-specific trunk reservation for stabilizing the throughput of the network and, on the other hand, a BIAS-related route-specific trunk reservation in order to take into account the progressive traffic load status of an entire alternative route in the grid of the updating interval ( decentralized, high-load-specific fixed or adaptive two-link travel reservation).

In reality, the number of free trunks of a link stored in the local trunk status table according to FIG. 2 is not already reduced by a certain number of trunk. Rather, the aforementioned artificial reduction takes place through the routing process for the purpose of determining the alternative route sequence.

The BIAS-related trunk reservation mentioned can also be changed in real time depending on the traffic load on a link. In order to determine the real-time BIAS value, the estimate of the real-time link-specific BIAS value BIAS (t) over the prediction interval dT is used at the end of each prediction interval (approximately equal to the update interval dT) educated:

BIAS _i (t, dT) = dT (a _i (t, dT) -

) (1)

where a. (t, dT) the call arrival rate, x. (t) the number of calls

TIMES ^r t oint currently occupied trunks of a link i, 't "m, the average call duration and ^χ * (t) / t is the Verbindungsbe- endigungsrate are. These BIAS estimates are themselves random variables. In particular, the call arrival rate itself according to a. (T, dT) = a. (T-dT, dT) + (1 -) Z. (dT) / dT (2) with 0 p 1 by moving averaging from the number of call arrivals Z. (dT) can be obtained in the prediction interval dT. For ß = 0, for example, almost instantaneous but varied estimates result, while ß = 1 leads to a _i (t, dT) = a _i (0, dT), that is, updates the initial value.

The values S = 0.9 and t = 180 s are often used.

A purely arithmetical calculation of expected values via equation (1) provides an illustration

E (BIAS. (T, dT)) = ■ dT! (A. (t) - y. (T)) (3)

where A. (t) is the traffic offer and y. (t) the load on link i at time t. This formation of expectations is real-time, i.e. for dT «t, not realizable, but can be used to estimate fixed BIAS reference values in experimental load situations. Amounts e.g. the current average bundle load y = 80 Erl, the currently incoming mean offer A = 116 Erl and the ratio dT / tm = 0, '10, 'would be close to real time

To set the BIAS value for the time interval from t to (t + dT) equal to 3.6.

In live operation, the smooth updates of a. (t) and the current status information x. (t) for an appropriate BIAS adaptation, which can be adjusted, for example, by means of the parameters dT and ß so that the processing load remains limited at very high loads. Since the prediction error increases with increasing prediction interval dT, this should be chosen as small as possible, but larger than all possible round-trip delays in the signaling system, if a cooperative import of status information for the completion of the local one, which will be explained in more detail below Trunk state table LTSM is applied.

The information about the availability of an intermediate switching node TN and the states of a second link L2 are imported after each update interval by the corresponding intermediate switching nodes via the character channels and stored in the local trunk status tables. In the case of a communication network with N switching nodes, each additional switching node must import and store additional second-link status messages from (N1) destination switching nodes in the route update interval dT = 10 s from the specified number of alternative routes. With N = 64, M = 14 and 2 bytes per state of a link, the database for the second link would therefore have the size of 28 bytes for each destination and thus a total size of 1.764 k bytes per originating switching node for all destinations. For the message transfer on the central character channel, it can be assumed that the switching nodes continue to be coupled via associated 64 kbit / s character channels. The total arbitrate rate distributed over the route update interval for importing into or exporting from the local trunk status table is therefore 2 DB x 8 / dT = 2.822 kbit / s per switching node. However, messages are exchanged in MSU datagram mode. Each MSU (Message Signaling Unit) of gross nominal about 272 bytes find their destination within about 100 msec. For each MSU and target, M x 2 bytes = 28 bytes must be transferred within dT or around 0.1 MSU / s. Furthermore, the originating switching node requests all intermediate switching nodes to report back their load and destination bundle states. For this purpose, M bits are required for the information about the availability of the intermediate switching node.

As an alternative to the aforementioned polling method, the bundle states of the second link sections can also be brought about from the intermediate switching nodes. This again requires M (N - 1) 2 bytes. The scope of the report practically does not change.

The far-sighted routing method results from the cooperative transfer of the information from the local Truπk status tables between the switching nodes of the non-hierarchical communication network. This in turn means that the much more critical and costly call processing is relieved. At the same time, a load-sharing and safety-relevant route processing network with decentralized databases is created with a low communication volume. By comparison, with a corresponding method with a central routing process, the N-fold volume of the database within dT would have to be transferred to two security-relevant, spatially diversified dual computers, which in turn had the same time grid of N (N - 1), i.e. 2 generate approximately N link states of alternative route sequences with up to M alternative routes and in N decentralized

Table storage would have to be transferred back.

Instead of the cooperative transfer of the second link status information stored in the local trunk status tables between the switching nodes, a method for intelligent learning of the status of the corresponding links and switching nodes are used. This method could be implemented, for example, in such a way that the states mentioned are estimated and updated in larger time intervals on the basis of conditional blocking experiences (subsequent links blocked under the condition first links free). This variant is particularly suitable for international connections with pronounced medium-term multi-hour traffic profiles and limited exchange of status information between the originating and destination switching nodes, as well as more than two link sections between these switching nodes.

3 shows selected analytical case studies for symmetrical and dynamically not hierarchically controlled

Communication networks with sequential mutual overflow (dynamic non hierarchical routing dNHR). The symmetry includes full meshing, n = 100 trunks per link and the same traffic offers A / n in Erl per trunk and traffic pair, but different alternative path sizes M as the only remaining optimization parameter in the present symmetrical case. The most suitable value TR = 4 is used as the trunk reservation TR, and the BIAS values are neglected. The successful traffic throughput is shown, i.e. the normalized traffic load y / n in Erlang per trunk as a function of the normalized traffic offer A / n in Erlang per trunk.

For M = 14, an almost loss-free throughput is achieved up to the offer 0.88 Erl / Trunk. Without a drink reservation TR, the break point would be followed by a crescent drop in throughput. With higher traffic offers, however, the highest possible throughput values are only achieved if the scope of the alternative route fan, i.e. the parameter M is reduced.

For A / n> l Erl / Trunk the uncontrolled communication network dominates with M = 0, because otherwise it would the traffic directed to the second-link alternative routes each hinders the direct traffic of the corresponding links twice.

On the other hand, in the case of M = 0, all excess incoming calls for smaller traffic offers, despite the existing capacity reserves of on average (n - A) per link, did not have any further chances of success, so that higher blockages in the percentage range compared to an optimal waybox size M> 0 result.

The state-controlled dynamic non-hierarchical routing method (SDNHR), however, solves the adaptation problem mentioned not only for symmetrical, but also for asymmetrical networks. As stated, busy alternative routes are not used and, in addition, priority is given to those alternative routes that have the largest almost real-time capacity reserves. In addition, the danger of alternative route overselections is countered by adaptive BIAS predictions in the route update interval. Here, the throughput is optimized again, so that the end region shown by a first hatching S1 in FIG. 3 is approximated over all loads. This area is also approaching the ideal with increasing link capacities for n 100

Limit curve (y = A for A n or y = n for A ≥ n, indicated in FIG. 3 by a second hatching S2).

4 illustrates simulations on an asymmetrical network with N = 5 switching nodes and M = 3, with regard to an expected processing load

PRLR ratio for congested transport offers A in Erlang

Erl.

The Processing Load Ratio PRLR denotes the average call processing load per call due to sequential alternative path overflows and crankbacks. For planning or extreme loads, the Processing Load Ratio strives against 1, because in these cases practically no traffic overflows occur or are permitted. In these cases, only the planning route is checked for each call.

A first curve K 1 shows the PRLR curve of a sequential routing method without the involvement of a local trunk status table, the calculation and simulations matching at negligibly small confidence intervals.

Significant reductions in the processing load ratio according to a second curve K2 result in the relevant high-load range if the locally available state information of the first links is used to update the routing sequences (alternative path sequences). This ensures that the first link (first section of an alternative route) is free, so that only the crankback of the second link (second section of the alternative route) is received for the processing load ratio. .

The BIAS term fixed in the simulations according to a third curve K3 reduces the estimation errors for the prediction interval dT (dT = 10 s). This embodiment of the invention is of particular interest for extensive hardware implementation of route processing, because the existing call processing would not be affected.

If the occupancy states are also imported from the secondary links of the alternative routes, the processing load ratio is further reduced to exemplary maximum values of size 1.06. This fact is illustrated by a fourth curve K4 and a fifth curve K5 light, whereby the fourth curve without BIAS reservation and the fifth curve was determined on the basis of a BIAS value of 5. This embodiment of the invention is therefore particularly suitable for its extensive software implementation in existing hardware system environments.

Claims

1. A method for non-hierarchical traffic control in a communication network, which a) comprises a plurality of switching nodes which are interlinked via links (links) such that there are several connecting paths between two such switching nodes, accordingly b) the traffic between two such switching nodes via at least one Planned route is routed, c) the named traffic is routed according to an alternative route sequence (AWS) via alternative routes if a connection via a planned route is not possible, d) the alternative route sequence represents an order of alternative routes ordered according to increasing traffic load conditions and is determined periodically in real-time intervals , characterized in that e) the alternative route sequence is determined from a predetermined set of alternative routes, an alternative route, the traffic load state of which exceeds a certain threshold value, not in the alternative route sequence is included.

2. The method of claim 1, d a d u r c h g e k e n n z e i c h n e t that said alternative route sequence (AWS) is determined in the respective originating switching node.

3. The method according to claim 1 or 2, characterized in that the traffic load status of an alternative route is classified based on the occupancy status of the outgoing from the originating node first link of the alternative route.

4. The method according to any one of claims 1 to 3, so that the traffic load state of an alternative route with respect to the follow-up links is classified on the basis of blocking experiences.

5. The method according to claim 1 or 2, d a d u r c h g e k e n n z e i c h n e t that the traffic load state of an alternative route is classified based on the occupancy status of all connecting routes forming the alternative route and the availability of the intermediate switching nodes lying on the alternative route.

6. The method according to any one of claims 1 to 5, so that the determined threshold value is changed in a link-specific and real-time manner as a function of the difference between call arrivals and connection completions (call completion) averaged over a prediction interval.